Transcript Study design and efficiency

Ioannis Sarigiannidis & Diego Lorca
MfD – Wednesday 23rd January 2013
Ioannis Sarigiannidis & Diego Lorca
• Trials may consist of
– ‘events’ (burst of neural
activity)
– blocks’ (sustained neural
activity)
• ITI (intertrial interval): time
between the end of a trial and
the beginning of the next one
• SOA (stimulus onset
asynchrony): time between
the onset of components
PREDICTED ACTIVATION IN VISUAL AREA
BOXCAR
x
(convolve)
CONVOLVED
WITH HRF
General Linear Model:
Y
=
X
.
β
+
ε
Data
Design Matrix Parameters error
Efficiency(e) is the ability to estimate β, given the design
matrix (X) for a particular contrast (c)
e (c, X) = inverse (σ2 cT Inverse(XTX) c)
All we can alter in this equation is in fact X
-A signal that varies little will be difficult to detect.
Fixed SOA 16s
HRF
Fixed SOA 4s: low variance
HRF
Random SOA minimum 4s e.g. event-related: larger variability in signal
HRF
Blocked, SOA 4s: larger variability in signal
HRF
• All waveforms in the universe, are actually just
the sum of simple sinusoids of different
frequencies.
the one we want to decompose into its “building blocks”
first component
prediction
the one we want to decompose into its “building blocks”
second component
prediction
the one we want to decompose into its “building blocks”
third component
prediction
fourth component
prediction==initial waveform
HRF
Low pass filter!
HRF
•
fMRI noise tends to have two components:
– Low frequency ‘1/f’ noise
-physical (scanner drifts);
-physiological [cardiac (~1 Hz);
respiratory (~0.25 Hz)]
– Background white noise
•
SPM uses a high-pass filter
– Goal: maximise noise reduction & minimise loss of signal.
•
High-pass filter + low-pass filter (inherent) = ‘band-pass’ filter
HRF
x
frequency is lower
than high-pass
cutoff so, much
energy is lost
With randomised designs,
•optimal SOA for differential effect
(A-B) is minimal SOA (>2 seconds),
•optimal SOA for main effect (A+B) is
16-20s
Differential Effect (A-B)
Common Effect (A+B)
4s smoothing; 1/60s highpass filtering
•Introducing null events
 Efficiency for differential
and main effects at short SOA
• Blocked designs generally more efficient
– keep SOAs short
– do not exceed optimal block length
• Randomize the order, or randomize the SOA,
of conditions that are close in time.
Ioannis Sarigiannidis & Diego Lorca
Functional MRI is difficult.
Experiments on cognition are that much more
difficult because cognitive function is a moving
target.
Experimental design is not as self-consistent nor
as theory-based as most of the other aspects of
fMRI-based research.
(Savoy, 2005)
Paradigm
• Construction
• Temporal organization
• Behavioural predictions
Hypothesis
Scientific
question
Neuroanatomical basis
Suitable for Neuroimaging?
(Amaro & Barker, 2006)
Categorical
• comparing the activity between stimulus
types.
Parametric
• continuous values, and may have as many
‘levels’ as there are values (Henson, 2006).
Factorial
• combining two or more factors within a task
and looking at the effect of one factor on the
response to other factor.
FACTORIAL DESIGN
 Simple Main Effects
e.g. A-B = Simple main effect of Treatment
(vs. Control) in the context of Drug 1
 Main Effects
e.g. (A + B) – (C + D) = the main effect of
Drug 1 (vs. Drug 2) irrelevant of DRUG
Psychotherapy
 Interaction Terms
e.g. (A - B) – (C – D) = the interaction effect
of Treatment (vs. Control) greater under
Drug 1 (vs. 2)
PSYCHOTHERAPY
TREATMENT CONTROL
1
2
A B
C D
FACTORIAL DESIGN IN SPM
• Main effect of Drug 1:
• (A + B) – (C + D)
• Simple main effect of Treatment in the
context of Drug 1:
• (A – B)
A
B
C
D
[1
1
-1
-1]
A
B
C
D
[1
-1
0
0]
A
B
C
D
-1
1]
• Interaction term of Treatment greater under
Drug 1:
• (A – B) – (C – D)
[1
-1
Blocked
based on maintaining cognitive engagement in a task by presenting
stimuli sequentially within a condition, alternating this with other
moments (epochs) when a different condition is presented (Amaro
& Barker, 2006).
“on” Block “off” Block
(Petersen & Dubis, 2012)
Advantages:
• Robustness of results.
• Increased statistical power.
• Relatively large BOLD signal change
related to baseline.
Disadvantages:
• Induce differences in the cognitive ‘set’ or
strategies adopted by subjects.
• Cannot distinguish between trial types
within a block.
(Amaro & Barker, 2006; Henson, 2006; Petersen & Dubis, 2012)
Event-related
discrete and short-duration events are presented with timing and
order that may be randomized (Tie et al., 2009).
Slow eventrelated design
inter-stimulusinterval (ISI)
Rapid eventrelated design
longer
HRF
shorter
(Petersen & Dubis, 2012)
Advantages:
o Analyses related to individual responses to trials.
o Less sensitivity to head motion artifacts
o Can be used to assess practice effects.
o Randomization of the order of conditions presented.
o Can also vary the time between stimulus presentation.
o Post hoc methods.
o Maintenance of a particular cognitive or attentional set.
o Decrease the latitude the subject has for engaging alternative strategies.
Disadvantages:
 Reduced ability to estimate the HRF properties of a single stimulus (Rapid
event-related fMRI)
 Linearity versus non-linearity of the BOLD interaction in overlapping HRFs
(Rapid event-related fMRI).
 Decrease of signal-to-noise.
(Amaro & Barker, 2006; Friston, 2003; Petersen & Dubis, 2012)
(Tie, 2009)
Pre-surgical
mapping
Antonym
generation task
Up….Down
Remember jittering!!!
Blocked design VS
Event-related design
Mixed designs
A combination of block and event-related designs can provide information
relating to ‘maintained’ versus ‘transient’ neural activity during paradigm
performance.
baseline
Task block
events
Advantages:
o Item-related pattern of information processing (transient activity).
o Task-related information processing (sustained activity).
Disadvantages:
 Involves more assumptions than other designs.
 Issues associated with poorer HRF shape estimation.
 Post hoc analysis of behaviour correlated activation.
 Power considerations.
(Amaro & Barker, 2006; Petersen & Dubis, 2012)
(Donaldson et al., 2001)
fMRI Adaptation (Repetition Suppression)
fMRI responses
Pairs of stimuli
Identical
feature
VS
Stimulus pairs
Differ in the
feature
Reduced
responses
Identical pairs VS
nonidentical pairs
Neuronal
selectivity
Similar
responses
Repetitions and
non repetitions
Lack of
selectivity
(Larsson & Smith, 2012)
(Larsson & Smith, 2012)
Main Advantage:
• Potential superiority over standard fMRI paradigms with regard
to tapping into functional specificity of neuronal populations
(Segaert et al., 2013).
Resting state fMRI
An alternative to the paradigms used in conventional designs is
to let the subject lay inside the MR scanner doing nothing, and
observe variations of the BOLD response related to spontaneous
activity, or ‘resting state’.
Advantages:
• Method of functional connectivity.
Disadvantages:
• May have limitations related to linearity
properties of overlapping HRFs.
• Statistical power of the study is largely
unknown beforehand.
(Amaro & Barker, 2006; Lowe, 2012)
(Donaldson, 2004)
Think about your study EFFICIENCY,
CONTRASTS before you start!
DESIGN
and
•
•
•
•
•
•
•
•
http://imaging.mrc-cbu.cam.ac.uk/imaging/DesignEfficiency
http://www.thefouriertransform.com/
Previous MfD slides
Amaro, E., & Barker, G. J. (2006). Study design in fMRI: Basic principles. Brain
and Cognition, 60, 220-232. doi: 10.1016/j.bandc.2005.11.009
Donaldson, D. I. (2004). Parsing brain activity with fMRI and mixed designs:
what kind of a state is neuroimaging in? Trends in Neurosciences, 27, 442-444.
doi: 10.1016/j.tins.2004.06.001
Donaldson, D. I., Petersen, S. E., Ollinger, J. M., & Buckner, R. L. (2001).
Dissociating state and item components of recognition memory using fMRI.
Neuroimage, 13, 129-142. doi: 10.1006/nimg.2000.0664
Friston, K. J. (2004). Introduction: experimental design and statistical
parametric mapping. In R. S. J. Frackowiak et al. (Eds.), Human brain function
(pp. 599-634). London: Academic Press.
Henson, R. (2006). Efficient experimental design for fMRI. In K. J. Friston, J. T.
Ashburner, S. T. Kiebel, T. E. Nichols, & W. D. Penny (Eds.), Statistical
parametric mapping: The analysis of functional brain images (pp. 193-210).
London: Academic Press.
• Larsson, J., & Smith, A. T. (2012). fMRI repetition suppression: Neuronal
adaptation or stimulus expectation? Cerebral Cortex, 22, 567-576. doi:
10.1093/cercor/bhr119
• Lowe, M. J. (2012). The emergence of doing ''nothing'' as a viable paradigm
design. Neuroimage, 62, 1146-1151. doi: 10.1016/j.neuroimage.2012.01.014
• Petersen, S. E., & Dubis, J. W. (2012). The mixed block/event-related design.
Neuroimage, 62, 1177-84. doi: 10.1016/j.neuroimage.2011.09.084
• Savoy, R. L. (2005). Experimental design in brain activation MRI: Cautionary
tales. Brain Research Bulletin, 67, 361-367. doi:
10.1016/j.brainresbull.2005.06.008
• Segaert, K., Weber, K., de Lange, F. P., Petersson, K. M., & Hagoort, P. (2013).
The suppression of repetition enhancement: A review of fMRI studies.
Neuropsychologia, 51, 59–66. doi: 10.1016/j.neuropsychologia.2012.11.006
• Tie, Y., Suarez, R. O., Whalen, S., Radmanesh, A., Norton, I. H., & Golby, A. J.
(January 01, 2009). Comparison of blocked and event-related fMRI designs for
pre-surgical language mapping. Neuroimage, 47, T107–T115. doi:
10.1016/j.neuroimage.2008.11.020
Special thanks to our expert
Thomas Fitzgerald!